{"openapi":"3.1.0","info":{"title":"compute.md","version":"1.0.0","description":"The COMPUTE.md skill with your limits written in, plus read-only sizing estimates and accelerator specs. No key, no prices. Provided as is; information may not be current or accurate; use at your own risk.","license":{"name":"MIT","identifier":"MIT"},"termsOfService":"https://compute.md/terms"},"servers":[{"url":"https://compute.md"}],"paths":{"/compute.md":{"get":{"operationId":"computeSkill","summary":"The COMPUTE.md skill. Add any lever to get a version with a Your limits section written in. Every lever is a number or a fixed choice.","parameters":[{"name":"budget_usd","in":"query","required":false,"description":"Budget per task ($), 1 to 1000000.","schema":{"type":"number"}},{"name":"approval_usd","in":"query","required":false,"description":"Ask before any spend over ($), 0 to 1000000.","schema":{"type":"number"}},{"name":"max_hourly_usd","in":"query","required":false,"description":"Most per hour, all machines ($), 0.1 to 100000.","schema":{"type":"number"}},{"name":"model_budget_usd","in":"query","required":false,"description":"Model calls per task ($), 0 to 1000000.","schema":{"type":"number"}},{"name":"stop_hours","in":"query","required":false,"description":"Hard stop after (hours), 0.25 to 720.","schema":{"type":"number"}},{"name":"idle_minutes","in":"query","required":false,"description":"Shut down when idle for (minutes), 5 to 1440.","schema":{"type":"number"}},{"name":"checkpoint_minutes","in":"query","required":false,"description":"Checkpoint every (minutes), 5 to 240.","schema":{"type":"number"}},{"name":"spot","in":"query","required":false,"description":"Spot or interruptible","schema":{"type":"string","enum":["never","checkpointed","preferred"]}},{"name":"providers","in":"query","required":false,"description":"Providers","schema":{"type":"string","enum":["existing","ask"]}},{"name":"regions","in":"query","required":false,"description":"Regions","schema":{"type":"string","enum":["any","data","us","eu"]}},{"name":"approver","in":"query","required":false,"description":"Who approves","schema":{"type":"string","enum":["user","owner","requester"]}}],"responses":{"200":{"description":"The skill as Markdown","content":{"text/markdown":{"schema":{"type":"string"}}}},"400":{"description":"Bad parameter, with a hint"},"404":{"description":"Model not found"}}}},"/api/v1/fit":{"get":{"operationId":"fitModel","summary":"Size a model or job: memory, which GPUs fit and how many, and estimated speed. Estimates only; no prices.","parameters":[{"name":"model","in":"query","required":false,"description":"Hugging Face repo id (Qwen/Qwen3.6-27B) or a quick pick (gpt-oss-120b). Or pass params_b instead.","schema":{"type":"string"}},{"name":"task","in":"query","required":false,"description":"Default serve.","schema":{"type":"string","enum":["serve","lora","qlora","full"]}},{"name":"precision","in":"query","required":false,"description":"Weight precision. Default native: the checkpoint as published.","schema":{"type":"string","enum":["native","bf16","fp8","int8","int4"]}},{"name":"kv","in":"query","required":false,"description":"KV cache precision. Default bf16.","schema":{"type":"string","enum":["bf16","fp8"]}},{"name":"context","in":"query","required":false,"description":"Tokens per sequence, such as 32768 or 32k. Default 32k serving, 4k training.","schema":{"type":"string"}},{"name":"concurrency","in":"query","required":false,"description":"Sequences in flight when serving, batch size when training. Default 8 or 4.","schema":{"type":"integer"}},{"name":"train_tokens","in":"query","required":false,"description":"Total training tokens (dataset x epochs), such as 50m, for an estimated duration.","schema":{"type":"string"}},{"name":"max_gpus","in":"query","required":false,"description":"Largest device count to consider. Default 8.","schema":{"type":"integer"}},{"name":"gpu","in":"query","required":false,"description":"Only these accelerators, comma-separated: h100-sxm,h200,mi300x.","schema":{"type":"string"}},{"name":"limit","in":"query","required":false,"description":"Options to return. Default 12.","schema":{"type":"integer"}},{"name":"params_b","in":"query","required":false,"description":"Parameters in billions, for a model that isn't on Hugging Face. Add active_b, layers, kv_heads and head_dim for speed and KV cache.","schema":{"type":"number"}}],"responses":{"200":{"description":"JSON result","content":{"application/json":{"schema":{"type":"object","additionalProperties":true}}}},"400":{"description":"Bad parameter, with a hint"},"404":{"description":"Model not found"}}}},"/api/v1/gpus":{"get":{"operationId":"gpuSpecs","summary":"Accelerator specs from vendor datasheets: memory, bandwidth, dense BF16 and FP8 throughput, interconnect. No prices.","parameters":[{"name":"gpu","in":"query","required":false,"description":"Accelerator ids or names, comma-separated: h100, h200, b200, a100, l40s, mi300x, rtx-4090, tpu-v6e.","schema":{"type":"string"}},{"name":"min_memory_gb","in":"query","required":false,"description":"Only accelerators with at least this much memory.","schema":{"type":"number"}},{"name":"vendor","in":"query","required":false,"description":"Only one vendor: NVIDIA, AMD, Google, AWS, Intel, Apple.","schema":{"type":"string"}}],"responses":{"200":{"description":"JSON result","content":{"application/json":{"schema":{"type":"object","additionalProperties":true}}}},"400":{"description":"Bad parameter, with a hint"},"404":{"description":"Model not found"}}}},"/api/v1/models":{"get":{"operationId":"modelSpecs","summary":"Model specs: parameters, active parameters, checkpoint size, attention layout and KV cache per token. Quick picks, or any Hugging Face repo via id.","parameters":[{"name":"id","in":"query","required":false,"description":"Hugging Face repo id. Omit for the quick-pick list.","schema":{"type":"string"}}],"responses":{"200":{"description":"JSON result","content":{"application/json":{"schema":{"type":"object","additionalProperties":true}}}},"400":{"description":"Bad parameter, with a hint"},"404":{"description":"Model not found"}}}}}}